10/24/2020

Motivation

  • Social network has great impacts on language learning in the study abroad context (Dewey 2017; Gautier, 2017; Isabelli-García, 2006; Shiri, 2015).
  • Ego-centric network studies suggest that study abroad students lack critical access to out-of-classroom social networks and thereby they tend to interact with other students in the same cohort (Dewey, 2013; Shiri, 2015).
  • Little work has been done to examine the internal interaction of study abroad students (Paradowski et al. 2020).
  • Given that prior studies focus on ego-centric networks, it also fails to examine how broad network structure shapes students’ language learning.

Our solution

  • We developed a novel Study Abroad Complete Network Questionnaire to collect students’ complete or whole network data.
  • We collected SACNQ in a small-size (38) study abroad program in China.

Here is a preview of SACNQ.

  • We first obtained a list of students in the target cohort we were studying.
  • We then asked study abroad students to identify each student they frequently talk to in their cohort.
  • We further asked students to evaluate each selected student’s language use.
  • We also collected each student’s demographic and other friendship network information.
  • We further conducted a 6-month ethnographic work to assess the validity of students’ self-reported network from SACNQ.
  • Note that we asked students to report their out-of-classroom networks as well.

What do we have in SACNQ?

  • Ego-centric network approach
    • List up to a limited number of persons students frequently talk to
    • Incomplete network
  • Complete network approach
    • Strength: collect information for each pair of students in the cohort
    • Weakness: difficult to scale-up (appropriate for small group)

Here is a snapshot of network data

Here is a snapshot of network data

load("sacnq.RData")
knitr::kable(sacnq_edges[1:2,1:8])
from to w1_freq w2_chn w3_en w4_oth_lang w5_chn_pro w6_en_pro
1 2 1 5 95 NA 3 5
1 5 2 5 95 NA 2 5
knitr::kable(sacnq_attrs[1:2,c(1,3:8)])
name_id pseudo_name gender nation region month chn_pro_self_report
1 Tiffany F USA North America 10 Imd
2 Hana F Korea Asia 22 Imd

Let us visualize the speaker network first.

Let us see the internal community in the cohort.

Let us see the popularity of students.

Let us explore SA students’ language use.

Let us take a deep look at Chinese Net.

Do network structure and students’ attributes influence language use?

  • We can use exponential random graph model (ERGM) to estimate the Chinese speaking network formation.

Here is the ERGM output.

What does ERGM tell us?

  • SA students with near native or native level tend to speak with other cohort students with high level of Chinese.
  • SA students staying in China for at least half year are also more likely to use Chinese in their interaction with other students.
  • Female SA students tend to use Chinese in their interaction.
  • Female SA students also tend to talk with female students using Chinese.
  • Caucasians are less likely to use Chinese.
  • SA students with on-campus activities are less likely to use Chinese interacting with their cohort members.
  • SA students with off-campus activities tend to use Chinese when they interact with other cohort members

We use ethnographic work to validate our SACNQ.

  • Ethnographic data show similar pattern.

Some Caveats and Challenges

  • Data collection may encounter unexpected problems.
  • Difficult to Scale-up.